proompts

A curated set of prompts in YAML for AI-assisted product development, regulatory workflows, and general operations. Prompts are organized by topic: ranging from code reviews to market research. You can mix and match them in your own agentic workflows!

🗺️ Directory Map

Path	Description	Key Contents
prompts/	The Core Library. YAML-based prompts organized by domain.	business/, clinical/, technical/
workflows/	Chained Sequences. Multi-step prompt chains for complex tasks.	paw_workflow.workflow.yaml
tools/	The Engine Room. Utilities for maintenance and development.	scripts/ (Python), prompt_tools/ (Meta-prompts)
scripts/	Shell Wrappers. Convenience scripts for common tasks.	validate_prompts.sh
docs/	Knowledge Base. Guides, usage instructions, and generated docs.	USAGE.md, BEST_PRACTICES.md

Docs

• Full Table of Contents – Browse all prompts by category
• Usage Guide – Learn how to use and integrate prompts
• Developer Scripts – Tools for validating and maintaining the repo

Setup

To run validation scripts and tools locally, you need Python 3 and the required dependencies.

1. Create a virtual environment (optional but recommended):

   python3 -m venv venv
   source venv/bin/activate

2. Install dependencies:

   pip install -r requirements.txt

3. Run the validation script to verify everything is set up correctly:

   ./scripts/validate_prompts.sh

Prompt Schema

Prompts are stored as .prompt.yaml or .prompt.yml files. Each prompt file contains two sections:

• Runtime information – ordered messages with role/content pairs that form the actual prompt. Use {{variable}} placeholders for user-provided values.
• Development information – optional metadata that describes the prompt and how to test it.

Top-level fields available in a prompt file include:

• name – short human-readable title
• description – concise summary of what the prompt does
• model – model identifier
• modelParameters – optional model parameters such as temperature
• messages – list of system and user messages
• testData – example inputs with their expected outputs
• evaluators – rules for verifying model responses

See docs/template_prompt.prompt.yaml for a filled-out example.

Prompt Workflows

In addition to individual prompts, this repository supports Prompt Workflows, which chain multiple prompts together to perform complex, multi-step tasks. Workflows are defined in .workflow.yaml files and call prompts located in dedicated workflow subfolders (e.g., prompts/clinical/protocol/protocol_workflow/). included runner script.

To learn more, see the Prompt Workflows Documentation.

Validation

To run all validation checks (YAML linting, schema validation, documentation index verification) locally, use the provided script:

python3 tools/scripts/test_all.py

This script runs the following checks:

• cleanup_mac_files: Removes macOS metadata files (._*) to ensure a clean state.
• check_prompts: Verifies file naming conventions and directory structure.
• validate_prompt_schema: Ensures prompts follow the required schema (e.g., messages, testData).
• update_docs_index: Checks if the documentation index is up-to-date.
• generate_docs: Generates category and workflow documentation pages.
• check_broken_links: Scans for broken internal links in documentation.
• yamllint: Lints YAML files for formatting.

It is recommended to run this script before committing changes.

Contributing

See CONTRIBUTING.md for detailed instructions on how to contribute.

1. Create prompts as .prompt.yaml files that follow docs/template_prompt.prompt.yaml and place them in the appropriate folder.
2. Review the Best Practices Guide for detailed guidance on creating high-quality prompts.
3. Ensure your prompt includes:
   • Meaningful testData with realistic examples (at least 1-2 test cases)
   • evaluators to validate output quality
   • Clear instructions and expected output format
4. Before committing, run validation:

   ./scripts/validate_prompts.sh

5. If you create a new directory, an overview.md will be generated automatically by the workflow.

The same validation runs in CI, but running checks locally helps catch issues early.

License

This project is licensed under the Proompts Personal Use License. Individuals may freely use, modify, and distribute the prompts for personal, non-commercial purposes. Commercial entities must obtain written permission from Frederick de Ruiter before using the material.